CN114463028A

CN114463028A - Alternative merchandise boosting service

Info

Publication number: CN114463028A
Application number: CN202110727726.5A
Authority: CN
Inventors: 伊塔马尔·大卫·莱瑟森; 罗丹·丘丁; 朱莉·德沃拉·凯兹奥哈顿; 莫西·沙哈鲁
Original assignee: NCR Corp
Current assignee: NCR Voyix Corp
Priority date: 2020-10-30
Filing date: 2021-06-29
Publication date: 2022-05-10
Also published as: EP3992896A1; US20220138830A1; US11922478B2

Abstract

A commodity vector representing a transaction context for the commodity is mapped to the multidimensional space. A request for an alternative to a given good is received from a resource. The multidimensional space is evaluated based on respective commodity vectors to identify a candidate commodity that is closest to the given commodity. Selecting a best candidate item from the candidate items based on the request. Injecting an association between the given commodity and the best candidate commodity within a process workflow associated with the resource.

Description

Alternative merchandise boosting service

Technical Field

Background

Maintaining successful private brands is critical to retailers increasing profits, creating more loyalty customers, and increasing store diversity. However, retailers currently have no automated tools and services to promote their private brands, and therefore they often have to rely on the manual effort to drive their private brand sales.

In fact, it is difficult for retailers in the industry to achieve private brand sales, as this requires a great deal of effort from different departments including marketing, sales, advertising campaigns, finance, etc. When dealing with private brands, retailers need to ask themselves: 1) how to effectively promote private brand goods; 2) how private brand goods may be recommended as an alternative to natural brand goods; and 3) who is the direct competitor of each of the personal branded goods?

An increasing number of retailers are aware of the importance of their private brands-a global study has shown that 65% of shoppers consider private brands to be a good alternative to well-known brands. This result is much higher for the new millennium generation and the Z generation. Thus, retailers who fail to adopt effective strategies to promote and maintain their private brands will not be able to maintain competition (especially in the ever-evolving grocery market).

Retailers have low profit margins while selling nationwide branded goods, but have much higher profit margins for private branded goods. Thus, if a retailer can increase the sales of a private brand of merchandise, the retailer can increase profitability and competitiveness.

Disclosure of Invention

In various embodiments, methods and systems for an alternative merchandise boost service are presented.

According to one aspect, a method for alternative merchandise boosting services is presented. For example, a given merchandise code associated with a shopping basket currently transacted online is received, and a similarity value of the potential alternative merchandise code to the given merchandise code is obtained. A selected alternative merchandise code is selected from the potential alternative merchandise codes and provided to a transaction service associated with the current online transaction for recommendation as an alternative to the given merchandise code within the shopping basket.

Drawings

FIG. 1A is a diagram of a system for alternative merchandise boosting services, according to an example embodiment.

FIG. 1B is a diagram representing a process flow of the system of FIG. 1A, according to an example embodiment.

FIG. 2 is a diagram of a method for alternative merchandise boosting services, according to an example embodiment.

FIG. 3 is a diagram of another method for alternative merchandise boosting services, according to an example embodiment.

Detailed Description

Fig. 1A is a diagram of a system 100A for alternative merchandise boosting services, according to an example embodiment. It should be noted that the components are schematically shown in a greatly simplified form, in which only those components relevant for understanding the embodiments are depicted.

Moreover, various components (identified in fig. 1A and 1B) are illustrated and an arrangement of these components is presented for purposes of illustration only. It should be noted that other arrangements with more or fewer components are possible without departing from the teachings of alternative merchandise boosting services presented herein and below.

System 100A provides a mechanism by which merchandise associated with a retailer's private brand product may be enhanced via facilitating visibility and recommendation of the private brand product. The boost service provides an automatic mechanism that is independent of human intuition and manual effort. The boost service is based on real observations from past recorded transactions, rather than speculations. The boosting services are integrated within the retailer's existing systems, such as up-sell services associated with promotions, coupons, e-commerce, retailer product placement systems, mobile shopping applications, and the like. The "Word 2 Vec" algorithm is used to determine the transaction context for the commodity, and a commodity vector is generated representing the transaction context. Transactions in the transaction history are evaluated, each transaction representing a sentence, and each transaction including words represented by a commodity code for the commodity. The transaction context for each commodity code is processed by the Word2Vec algorithm to generate the corresponding commodity for the multidimensional space. Commodity vectors are plotted in the multidimensional space, and a similarity between commodities is determined based on a distance between the plotted commodity vectors within the multidimensional space. The personal brand good may be determined to be an alternative to the natural brand good based on a distance between the plotted natural brand good vector and the personal brand good vector within the multi-dimensional space. Alternative private brand goods may then be identified and used as recommended replacement or alternative goods for a given brand good for customers and retail personnel via integration with existing retail services, existing retail applications, and existing retail systems.

System 100A includes enterprise data repository 110, online trading system 120, in-store trading terminal 130, user device 140, inventory and commodity placement system 150, management application (app)160, commodity space mapper and similarity manager 170, one or more Machine Learning Algorithms (MLAs) 180, and alternative commodity manager 190. System 100A includes various computing devices, each including at least one processor and a non-transitory computer-readable storage medium including executable instructions. The executable instructions, when executed by a corresponding processor from a corresponding non-transitory computer readable storage medium, cause the processor to perform the operations discussed herein and below with respect to the

component

110 and 190.

"item code" means an item from a given retailer's product catalog. The similarity/likeness of each item code to other item codes defines the vector of the item in the multidimensional space. The similarity/similarity and commodity code vectors are determined by the Word2Vec algorithm based on an analysis of the retailer's transaction history and product catalog. A "commodity vector" is a mathematical expression that indicates the point in a multidimensional space that represents the context of a given commodity.

Enterprise data store 110 includes various enterprise data, such as transaction histories for transactions made with retailers. Other types of data may also be included in enterprise data store 100, such as incentives available to customers, customer data (loyalty data, profile data, etc.) for known customers, transaction details for customer transactions (including merchandise codes for merchandise), merchandise or product catalog data, and other information captured and retained by the enterprise for stores and other stores associated with the enterprise (retailer).

The online transaction system 120 includes an interface and corresponding software through which a customer conducts an online transaction with a retailer, for example, by browsing items, storing selected items in a virtual shopping cart, and settling (paying) in the virtual shopping cart. The online transaction system 120 may be web-based and/or mobile app-based. During the online shopping session, virtual shopping cart data is provided from the online trading system 120 to the enterprise data repository 110 in real-time or to the alternate merchandise manager 190 via an Application Programming Interface (API) in real-time when the selected merchandise of the virtual shopping cart is in inventory.

The transaction terminal 130 includes peripheral devices (scanners, printers, media receivers/dispensers, weighing devices, Personal Identification Number (PIN) pads, card readers, etc.) and corresponding software for performing customer settlement associated with the transaction. Real-time merchandise and transaction data is provided by the terminal to enterprise data repository 110.

The user device 140 includes a peripheral device (touchscreen, camera, etc.) and corresponding software for performing customer transactions using a web browser or mobile application (app). Real-time transactional data is provided by any app to enterprise data store 110.

The inventory and item placement system 150 includes means and corresponding software and user interfaces for performing inventory management and store layout and item placement (planogram). Any item code for an item being processed by the system 150 may be sent via the API to the alternative item manager 190 for an alternative item, such as a private brand item code.

Commodity space mapper and similarity manager 170 initially uses the transaction history record (obtained from enterprise data store 110) to generate a vector of commodity codes in the product catalog (likewise, obtained from enterprise data store 110). In this manner, the commodity codes in the product catalog are assigned vectors that map to the multidimensional space. Each vector is linked to the item code of the product catalog. The "goods embedding" is applied using the "Word 2 Vec" algorithm. Word2Vec is a set of algorithms used primarily for machine translation in the field of Natural Language Processing (NLP). Word2Vec takes as input a large corpus of text (product catalog of commodity codes and transaction history from enterprise data repository 110) and produces a vector space (multidimensional space) typically having hundreds of dimensions, with each unique Word (commodity code) in the corpus assigned a corresponding unique vector drawn in the multidimensional space. In this manner, merchandise codes that share a common context in the transaction history are drawn next to each other within the multidimensional space. The transaction history is provided to the Word2Vec algorithm as a sentence and the words are commodity codes (all available words are identified from the product catalog). Mathematical calculations may be applied to the vector numeric representation (vector) of the commodity code.

Once the product catalog and transaction history are processed to create a commodity code vector for the commodity, the commodity space mapper and similarity manager 170 may be provided with a given commodity code (as input or as a request for a suggested commodity replacement code) that represents a nationwide brand commodity obtained from any of the

components

120 and 160. The output produced by the Word2Vec algorithm is one or more alternative goods (private brand goods) and a similarity score (which corresponds to the distance between the location of the goods code of the nationwide brand goods and the location of the private brand goods within the multidimensional space). The goods space mapper and similarity manager 170 may determine the particular suggested alternative goods code to be provided to the alternative goods manager 190 based on a preset threshold or range of values and/or based on a predefined number of top similarity scores (the preconfigured number of highest similarity scores from the sorted list of similarity scores).

The goods space mapper and similarity manager 170 provides the generated alternative goods codes for the national brand goods codes to the alternative goods manager 190 along with the similarity scores. The alternative merchandise manager 190 provides the alternative merchandise code most likely to be purchased back to the original requester (component 120-160). Component 120-.

When the system 100A is deployed, feedback is monitored for actual transactions that are provided with alternative private brand merchandise codes for a given national brand merchandise code. The feedback is an indication of whether the alternative private brand goods code was purchased by the customer for the given transaction. The purchase is positive feedback and the non-purchase is negative feedback.

The one or more MLAs 180 are trained based on input including national brand commodity codes, private brand commodity codes, similarity scores (provided by the commodity space mapper and similarity manager 170 for any given national brand commodity code), and optionally customer vectors for customers associated with training transactions. The trained MLA180 represents a machine learning model for a particular application based on commodity similarity. Thus, as used herein, the MLA180 may also be referred to as a machine learning model.

In a similar manner as discussed above, the transaction history of the customer may be provided as sentences to another Word2Vec algorithm, and all the generated vectors are aggregated into a single customer vector representing the transaction history and preferences of a particular customer, which may be plotted in a multidimensional space along with a merchandise vector for identifying the relationship of the particular customer to the merchandise vector within the multidimensional space. The customer vectors are generated by the goods space mapper and similarity manager 170.

The training result that the MLA180 configures to achieve based on the input parameters provided is the selection of a particular alternative private brand commodity code, where the feedback indicates that a given customer actually purchased for a given transaction.

Once the MLA180 is trained, the alternate merchandise manager 190 receives the national brand merchandise codes for a given transaction in real time. The goods space mapper and similarity manager 170 passes the respective similarity scores between the alternative private brand goods codes and the national brand goods codes back to the alternative goods manager 190. The alternative merchandise manager 190 provides the customer vector, national brand merchandise codes, alternative private brand merchandise codes, and similarity scores for the customer associated with a given transaction as input to the MLA180 and receives as output the particular private brand merchandise codes that are most likely to result in a purchase of the particular customer's transaction. After the transaction, feedback is received as an indication of whether the particular proposed alternative private brand article was purchased by a particular customer of the transaction.

The MLA180 is continuously trained and updates the commodity code vector and the commodity code at configured intervals using updated product catalogs, new transaction data, and feedback. This ensures that the replacement private brand of goods for a particular nationwide brand of goods is tailored to the product catalog and the particular customer of a given store. In this way, the accuracy and success of suggested replacement/replacement goods will be in constant advancement and learning.

The

component

170, 190 can be provided to the retailers as a web-based and/or cloud-based service, with an API provided to the service to access the enterprise data store 110 of each retailer and communicate the alternative private brand goods code during the transaction. The API permits the management app 160 to be used by managers/retailers to obtain alternative private brand goods for nationwide brand goods on the retailer's shelves, ensuring that the shelf stock has the alternative private brand goods most likely to generate sales revenue for the retailer based on the transaction data of the given retailer, the given retailer's customer base, and the given retailer's particular product catalog.

The user-operated device 140 may be any customer-operated device, such as a wearable processing device, a voice-enabled network appliance (internet of things (IoT) device), a laptop, a desktop, a tablet, a network-based vehicle integrated device, and so forth. The device 140 may also be operated by retail personnel associated with the inventory and item placement system 150. The device 140 utilizes a retailer-provided interface (web-based and/or app-based) to perform shopping and transaction basket settlement with the transaction service of the web server 120.

The transaction terminal 120 may be a point of sale (POS) terminal, a self-service terminal (SST), an employee operated mobile device, and/or a kiosk.

The system 100A may be integrated into existing retailer applications, existing retailer services, and workflows of existing retailer systems. For example, the system 100A may be integrated into a workflow associated with boosting the purchase of private brand goods in a customer shopping basket (checkout goods or virtual shopping cart). Here, real-time recommendations during a transaction are made to a customer to replace certain nationwide branded goods present in the customer's shopping basket with alternative private branded goods by providing a better price or discount for the alternative private branded goods during checkout or while the customer is shopping online. For example, a customer operates a user device 140 having a retailer shopping application (mobile or browser-based) to shop. As the customer adds merchandise to his/her shopping basket or virtual shopping cart, the merchandise codes are reported to the alternative merchandise manager 190 in real time. The alternative merchandise manager 190 interacts with the merchandise space mapper and similarity manager 170 and the MLA180 and obtains corresponding alternative private brand merchandise that most likely causes the customer to replace nationwide brand merchandise within the shopping basket for purchase. The alternative merchandise manager 190 provides the alternative private brand merchandise code and shopping basket national brand merchandise code to the retailer application associated with the customer shopping via the device 140 via the API. The business rules are processed by the retailer application upon receiving two item codes that cause the retailer application to make real-time suggestions to the customer to replace nationwide branded items with alternative private branded items, along with an indication by doing so that the customer may save money and/or may obtain alternative private branded items at a price lower than the listed price using a discount.

As another example, an existing product recommendation engine may be integrated with the alternate merchandise manager 190 via an API call such that the at least one recommended merchandise may include alternate private brand merchandise. These recommendation engines may be biased or weighted to favor alternative private branded goods. For example, if the customer is going to

Pasta and olive oil were added to their shopping baskets. The recommendation service may typically recommend different brands of tomato sauce for subsequent addition by the customer. The alternative goods manager 190 dynamically pushes (via interaction with 170 and 180) the alternative private brand of tomato sauce via the API, causing the recommendation engine to give the private brand a higher weight in the recommendation list for the customer.

In yet another example, the system 100A is integrated with the inventory and item placement system 150 of a retail store to assist store sales clerks in passing next to itIt is popular to place private brand merchandise on store shelves for nationwide branded merchandise. For example, the salesperson uses the software of system 150 to assist them in placing products on the relevant shelves. When the salesperson places the goods on the pasta shelf, the system 100A will recommend brands of pasta next to what is considered most popular (based on analyzing millions of shopping baskets)

The pasta items place a personal brand pasta item.

Again, the MLA180 includes a learning mechanism. The user's response to the offer of the alternative private brand goods will be recorded and provided as feedback back to the alternative goods manager 190, which trains the MLA180 so that the recommendations fine tune over time. If the customer reacts positively to the recommendation, it will be marked as positive, otherwise it will be marked as negative.

FIG. 1B is a diagram representing process flow 100B of the system of FIG. 1A, according to an example embodiment.

FIG. 1B shows a more granular view of some of the components associated with system 100A.

The transactional data manager 111 provides transactional data from the enterprise data repository to the commodity space mapper and similarity manager 170. The item space mapper and similarity manager 170 generates a multidimensional vector space and a unique vector for each item code of the item catalog 112 drawn within the space.

The alternative merchandise manager 190 trains the MLA180 based on the inventory code, the alternative private brand merchandise code, the national brand merchandise code, the similarity values provided by the merchandise space mapper and similarity manager 170, and optionally the customer vector for the customer obtained from the loyalty transaction data 113 and feedback obtained for each customer across multiple channels in which transactions are conducted or effected for the customer by the online transaction system 120, transaction terminal 130, user device 140, and/or management app 160.

Subsequently, when any given shopping basket/transaction for a customer (via the online transaction system 120, transaction terminal 130, or user device 140) or any given request for alternate private brand goods for a given nationwide brand good is received from the management app 160 or inventory and goods placement system 150, the alternate goods manager 190 requests the goods space mapper and similarity manager 170 to provide an alternate private brand goods code and similarity value for the nationwide brand goods code based on the vector space and goods code vectors. The alternative merchandise manager 190 provides the national brand merchandise codes, the alternative private brand merchandise codes, and corresponding similarity values and optionally customer vectors for the customers as input to the MLA 180. The MLA180 provides as output a selection from the alternative private brand goods code that is best suited for the customer or best suited for the retail store (based on actual transaction data). The alternate merchandise manager 190 uses the API to communicate the selected best alternate private brand merchandise code appropriate for the given situation to the respective requesting party (online transaction system 120, transaction terminal 130, user device 140, inventory and merchandise placement system 150, or management app 160). The positive or negative results are fed back to the alternative merchandise manager 190 via the API or the positive or negative results are derived by the alternative merchandise manager 190 from the final transaction data or sales data associated with the transaction or request. This feedback is used in subsequent training sessions of the MLA 180.

The system 100A drives online, in-store, and mobile shopping to suggest and recommend private brands for nationwide brands. Recommendations are not simply based on the most similar private brands, but rather on actual feedback from previous transactions used to train machine learning models to obtain accurate customer-specific recommendations based on a given customer context. Private brand recommendations are integrated into existing retailer applications, existing retailer systems, and existing retailer services.

In one embodiment, the

components

110 and 113 and 170 and 190 are provided as a single cloud-based surface to the

components

120, 150 and 160 via an API.

These and other embodiments will now be discussed with reference to fig. 2-3.

Fig. 2 is a diagram of a method 200 for alternative merchandise boosting services, according to an example embodiment. The software module implementing the method 200 is referred to as a "commodity boosting service". The merchandise boost service is implemented as executable instructions programmed and residing within memory and/or non-transitory computer-readable (processor-readable) storage media and executed by one or more processors of the device. The processor of the device performing the merchandise booster service is specifically configured and programmed to handle the merchandise booster service. The commodity boosting service has access to one or more network connections during its processing. The network connection may be wired, wireless, or a combination of wired and wireless.

In one embodiment, the device for performing the goods boosting service is a server. In one embodiment, the server is a cloud processing environment that includes multiple servers cooperating with each other as a single server. In one embodiment, the server is a Local Area Network (LAN) server.

In one embodiment, the merchandise boosting service is all or some combination of 170-190 and/or process flow 100B.

In one embodiment, the merchandise booster service performs the processing discussed above with respect to system 100A and process flow 100B.

At 210, the merchandise boosting service receives a given merchandise code associated with a shopping basket of a current online transaction between a retailer and a customer of the retailer.

At 220, the commodity boosting service obtains a similarity value of the potential alternative commodity code to the given commodity code.

In an embodiment, at 221, the merchandise boosting service identifies the similarity value as a distance between the merchandise code vector plotted in the multi-dimensional space of potential alternative merchandise codes and a given merchandise code vector associated with the given merchandise code. This may be done in the manner discussed above with respect to the Word2Vec algorithm, which uses the transaction history to generate the transaction context for each commodity code as a commodity code vector plotted in a multidimensional space.

At 230, the merchandise boosting service selects a selected alternative merchandise code to the transaction service associated with the current online transaction for recommendation as an alternative to the given merchandise code within the shopping basket of the current online transaction.

In an embodiment, at 231, the merchandise boosting service selects a highest value from the similarity values and associates the highest value with the selected alternative merchandise code.

In an embodiment, at 232, the commodity boosting service provides the given commodity code, the potential alternative commodity code, and the similarity value as inputs to a trained machine learning model associated with a trained machine learning algorithm.

In an embodiment of 232 and at 233, the merchandise boosting service provides the customer vector plotted in the multidimensional space of the customer associated with the current online transaction as further input to the trained machine learning model. This can be done by the following steps: the loyalty data store 113 associated with the customer is accessed and the complete transaction history of the customer is obtained, then the Word2Vec algorithm is processed to generate the transaction context vector for the customer plotted in the multidimensional space. This permits the selection of the selected alternative merchandise code to be customer-customized to the customer.

At 240, the merchandise boosting service provides the selected alternative merchandise code to the transaction service associated with the current online transaction for recommendation as an alternative to the given merchandise code within the shopping basket of the current online transaction.

In the 233 and 240 embodiments, at 241, the merchandise boosting service receives the selected alternative merchandise codes as output from the trained machine learning model.

At 242 and in embodiments at 241, the merchandise boosting service receives feedback indicating the results of whether a replacement merchandise associated with a replacement merchandise code was purchased in place of a given merchandise associated with a given merchandise code using a current online transaction.

In an embodiment of 242 and at 243, the merchandise boosting service uses the feedback and transaction details associated with the current online transaction during a subsequent training session in which the trained machine learning model is retrained based on the feedback.

In one embodiment, at 244, the merchandise booster service sends the given merchandise code and the selected alternative merchandise code to the transaction service using the API.

In an embodiment, at 245, the merchandise booster service injects the transaction identifier, the given merchandise code, and the selected alternative merchandise code for the current online transaction into a process workflow associated with the transaction service.

In an embodiment, at 250, the merchandise boosting service provides the selected alternative merchandise code to the merchandise placement service for stocking alternative merchandise associated with the selected alternative merchandise code within the store adjacent to or in place of the given merchandise associated with the given merchandise code.

In one embodiment, at 260, the merchandise boost service is processed (210) 250 into a real-time cloud-based service that is provided to the transaction service during the current online transaction and during other online transactions processed by the transaction service.

FIG. 3 is a diagram of another method 300 for alternative merchandise boosting services, according to an example embodiment. The software module implementing method 300 is referred to as a "private brand boosting service". The private brand boosting service is implemented as executable instructions programmed and residing within memory and/or non-transitory computer-readable (processor-readable) storage media and executed by one or more processors of the device. The processor executing the private brand boosting service is specifically configured and programmed to process the private brand boosting service. The private brand boosting service has access to one or more network connections during its processing. The network connection may be wired, wireless, or a combination of wired and wireless.

In one embodiment, the means for performing the private brand boosting service is a server. In one embodiment, the server is a cloud processing environment that includes multiple servers cooperating with each other as a single server. In one embodiment, the server is a LAN server local to the retail store.

In one embodiment, the private brand boosting service is all or some combination of 170, 190, process flow 100B, and/or method 200.

The private brand boosting service presents another (and in some aspects) enhanced processing perspective relative to that described above with respect to fig. 2.

At 310, a private brand boosting service receives a nationwide brand goods code associated with nationwide brand goods from a resource.

In an embodiment, at 311, a private brand boost service receives a nationwide brand merchandise code from an online transaction system during an online transaction. Here, the resource is a transaction service that is processing an online transaction.

In an embodiment, at 312, the private brand boosting service receives a nationwide brand goods code from the transaction terminal during a transaction at the transaction terminal. Here, the resource is a transaction manager that is processing transaction terminal transactions.

In one embodiment, at 313, the private brand boosting service receives a nationwide brand goods code from the administrative system. Here, the resource is a product or item placement service that provides item layout recommendations and item shelf location recommendations to store managers operating the user device 140 and interacting with the placement service.

At 320, the private brand boosting service identifies a private brand goods code associated with the private brand goods as an alternative to nationwide brand goods.

In an embodiment, at 321, the private brand boosting service processes a trained machine learning model associated with a trained machine learning algorithm to identify a private brand commodity code.

At 330, the private brand boosting service provides the national brand goods code and the private brand goods code to the resource.

In an embodiment, at 331, the private brand boosting service provides the nationwide brand goods code and the private brand goods code to the recommendation system to cause the resource associated with the recommendation system to bias the recommendation related to the nationwide brand goods code with the private brand goods code.

It should be appreciated that the description of software in a particular form (e.g., component or module) is merely to aid understanding and is not intended to limit the manner in which software implementing those functions may be structured or constructed. For example, while the modules are shown as separate modules, they may be implemented as homologous code, as individual components, some, but not all of these modules may be combined, or the functionality may be implemented in software constructed in any other convenient manner.

Further, while software modules are shown as executing on one piece of hardware, the software may be distributed across multiple processors or in any other convenient manner.

The above description is illustrative and not restrictive. Many other embodiments will be apparent to those of skill in the art upon reviewing the above description. The scope of the embodiments should, therefore, be determined with reference to the appended claims, along with the full scope of equivalents to which such claims are entitled.

In the foregoing description of the embodiments, various features are grouped together in a single embodiment for the purpose of streamlining the disclosure. This method of disclosure is not to be interpreted as reflecting an intention that the claimed embodiments have more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive subject matter lies in less than all features of a single disclosed embodiment. Thus, the following claims are hereby incorporated into the detailed description, with each claim standing on its own as a separate exemplary embodiment.

Claims

1. A method, comprising:

receiving a given item code associated with a shopping basket currently transacting online;

obtaining a similarity value of a potential alternative merchandise code to the given merchandise code;

selecting a selected alternative merchandise code from the potential alternative merchandise codes; and

providing a selected alternative merchandise code to a transaction service associated with the current online transaction for recommendation as an alternative to the given merchandise code within the shopping basket.

2. The method of claim 1, further comprising providing the selected alternative merchandise code to a merchandise placement service for stocking alternative merchandise associated with the selected alternative merchandise code within a store adjacent to or in place of a given merchandise associated with the given merchandise code.

3. The method of claim 1, further comprising processing the method as a real-time cloud-based service to the transaction service.

4. The method of claim 1, wherein obtaining further comprises identifying the similarity value as a distance between a merchandise code vector plotted in a multi-dimensional space of the potential alternative merchandise code and a given merchandise code vector associated with the given merchandise code.

5. The method of claim 1, wherein selecting further comprises selecting a highest value from the similarity values and associating the highest value with the selected alternative merchandise code.

6. The method of claim 1, wherein selecting further comprises providing the given commodity code, the potential alternative commodity code, and the similarity value as inputs to a trained machine learning model associated with a trained machine learning algorithm.

7. The method of claim 6, wherein providing the given merchandise code further includes providing, as further input, a customer vector plotted in the multidimensional space of a customer associated with the current online transaction to the trained machine learning model.

8. The method of claim 7, wherein providing the given commodity code further comprises receiving the selected alternative commodity code as output from the trained machine learning model.

9. A system, comprising:

at least one processing device having at least one processor configured to execute instructions from a non-transitory computer-readable storage medium;

the instructions, when executed by the at least one processor from the non-transitory computer-readable storage medium, cause the at least one processor to perform operations comprising:

receiving a request associated with an alternative to a first merchandise code associated with a first merchandise;

determining a potential alternative merchandise code associated with a potential merchandise for the first merchandise code based on a transaction context associated with the first merchandise and the potential alternative merchandise;

selecting a best potential alternative commodity code from the potential alternative commodity codes based on the request; and

injecting an association between the first commodity code and the best potential alternative commodity code into a workflow associated with the request.

10. The system of claim 9, wherein the workflow comprises an online transaction workflow that is processing transactions associated with a shopping basket or shopping cart having the first merchandise code, or a merchandise recommendation workflow that is providing recommendations with respect to the first merchandise code.

11. A method comprising any one or any combination of the features of claims 1-8.

12. A system comprising any one or any combination of the features of claims 9-10.